Music Hack Day SF 2013

in San Francisco, CA


In Progress

Lyrical: A lyrics-based song-finding paradigm

Over the years, many apps have tried different ways to expose users to new songs. Some apps, such as Pandora or iTunes Genius use sound-similarity algorithms to find new, interesting songs. Other apps like Spotify try to use the social idea that close friends likely have similar tastes in music. There are many other paradigms on the market, but I think there is room for 1 more, and that's the space Lyrical aims to take.

By analyzing the words and patterns of words in a song's lyrics, we can characterize songs and try to understand what kinds of messages a user likes to listen to. Then, we can search for other songs with similar messages and recommend a subset of those songs that match certain sound characteristics(e.g. the tempo of other songs the person recently listened to). Thus, the goal of Lyrical is to make predictions about the kind of content somebody wants to listen to and to use that information to better inform song recommendations.

The project may involve the following fields:
Data Gathering, Analysis, Machine Learning, Natural Language Processing, front-end(web), backend(dynamic webserver)

0 Favorites




We've joined the Mashery family. Read the announcement.